{
  "context": "Tests whether the agent evaluates each schema-v5 classification domain independently, fails closed on a deliberately stale schema-v4/scoring-v4 pattern profile, preserves explicit non-pattern guardrails, and keeps the current taxonomy useful without inventing history.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Legacy provenance fails closed",
      "description": "The report treats the supplied speaker-profile schema 4 / scoring schema 4 pattern payload as history-disabled under the installed policy-bound schema 5 / scoring schema 5 contract, regardless of its internally consistent legacy counts; it does not mistake schema-v4 occurrence compatibility or its readable non-pattern lane for permission to use legacy derived pattern claims",
      "max_score": 12
    },
    {
      "name": "No historical four-tier structure",
      "description": "The report uses one flat relevant-pattern list and does not emit Signature, Contextual History, New to You, or Shake It Up as speaker-history tiers",
      "max_score": 10
    },
    {
      "name": "Per-domain classification claims suppressed",
      "description": "Because the legacy profile has no classification-availability-v2 domains or policy stamp, the report does not repeat usage rates, mastery, strengths, underuse, combinations, modes, severity, or trend from it. It emits no New-to-You/[NEW] claim: that tier is exactly an authorized never_tried classification, never not_yet_observed, a raw zero, or this stale never_used_patterns array",
      "max_score": 13
    },
    {
      "name": "No catalog recurrence from raw rows",
      "description": "The report emits no catalog [RECURRING] label and does not treat shortchanged or bullet-riddled-corpse as habitual based on stale antipattern_frequency rows. Catalog recurrence labels relay guardrail-check.py recurring_antipatterns records, and this Phase 2 task has no such emitted record",
      "max_score": 13
    },
    {
      "name": "Current taxonomy remains useful",
      "description": "The flat strategy recommends relevant current-taxonomy techniques for this outline without presenting them as proven speaker history",
      "max_score": 10
    },
    {
      "name": "Contextual findings retained",
      "description": "At least one antipattern detection is labeled [CONTEXTUAL] — a current-outline risk found by the current taxonomy rather than a historical claim (e.g., the meme-heavy opening sequence)",
      "max_score": 12
    },
    {
      "name": "Actionable recovery warning",
      "description": "The report explains that the relevant classification domains are unavailable, recommends reprocessing the stale scoring generation as needed and regenerating schema v5, and says the bundled speaker-toolkit-default@1 policy applies automatically rather than asking Morgan to supply thresholds",
      "max_score": 10
    },
    {
      "name": "Explicit non-pattern recurring issue preserved",
      "description": "The report applies the long_context_ramp warning at its declared severity because it carries source_lane non_pattern, while keeping it separate from catalog-derived [RECURRING] labels and stale pattern history",
      "max_score": 10
    },
    {
      "name": "Actionable recommendations",
      "description": "At least 2 pattern recommendations include specific advice for how to apply them to THIS talk — not generic definitions of what the pattern is",
      "max_score": 10
    }
  ]
}